Speculation Suggests Anthropic Might Be Hiding a Claude 3.5 Pro Model

teortaxesTex · x · 2026-08-13

The thread speculates on Anthropic's model updates. A user questions why the company would release the exact checkpoint used to train 0731 two weeks later if no additional RL/OPD training was done. The author suggests that, referencing the precedent from V3.1 to 3.1-Terminus, these two weeks might have been used to polish a V4-Pro specifically for agents, rather than making general capability progress.

Original post →

More from Models

Models channel →